Evidential reasoning predicts aerial threats with 93% accuracy from partial data
88% classification from short data slices — no full time series needed.
A new paper on arXiv (2607.14606) introduces an evidential reasoning approach for aerial target classification and intent prediction that works from partial data. The method splits time-series into short sub-samples, runs classifiers on each, then fuses results via belief propagation and an evidential reasoning framework. This reduces the need for long-duration data, critical in fast-moving combat scenarios where decisions must be made quickly. Target intent is inferred using rule-based techniques and a distance-based combination method that fuses information over time.
Because no public dataset exists, the team generated their own labeled dataset for evaluation. In a case study with eight targets, the system achieved 88% accuracy for target type classification and 93% for intent prediction. The framework explicitly minimizes false predictions — a key requirement in high-risk environments where uncertainty is safer than a wrong assessment. This work could enable faster, more reliable tactical decisions for combat aircraft, adapting to evolving threats with limited sensor data.
- Uses short sequential sub-samples instead of full time series for faster classification
- Achieves 88% accuracy for target type and 93% for intent prediction across eight targets
- Evidential reasoning framework manages uncertainty, minimizing false predictions in combat scenarios
Why It Matters
Real-time aerial threat assessment from partial data could transform combat aircraft decision-making.